goneflash / my_vio

My visual (inertial) odometry experiments

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Usage

  1. git submodule update --init --recursive
  2. mkdir build && cd build
  3. cmake .. && make

If want to build tests, then use the following instead of 3) : cmake .. -DUSE_TEST=ON && make

Feature Tracker App:

./feature_tracker/feature_tracker_app -p ~/Project/vio/data/desk2_subset/ -f jpg --detector FAST(SURF, ORB, ...) --descriptor DAISY(ORB, SIFT, ...) --matcher_type GRID(OCV)

##Log:

10/02/2017

Now set default to use FAST and FREAK. ./build/feature_tracker/feature_tracker_app -p feature_tracker/test/test_data/long_seq/ -f jpg

03/28/2017

Now run ./feature_tracker/feature_tracker_app -p ../feature_tracker/test/test_data/long_seq/ -f jpg --detector FAST --descriptor DAISY --matcher_type GRID ./feature_tracker/feature_tracker_app -p ../feature_tracker/test/test_data/long_seq/ -f jpg --detector FAST --descriptor DAISY --matcher_type OCV ./feature_tracker/feature_tracker_app -p ../feature_tracker/test/test_data/long_seq/ -f jpg --detector ORB --matcher_type GRID

The difference is OCV matching has more long feature tracks than grid search method. Maybe increase the search region? Although grid is much faster than OCV TODO: Evaluate how to decide use which method.

##OSX CMake not use homebrew export CMAKE_PREFIX_PATH=/usr/local

About

My visual (inertial) odometry experiments


Languages

Language:C++ 91.5%Language:CMake 7.6%Language:Shell 0.9%